Local Illumination Compensation in Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding technologies face challenges in achieving optimal compression ratios and complexity reduction, particularly in parallelized implementations, and lack effective methods for local illumination compensation.

Innovation Solution

The integration of local illumination compensation (LIC) techniques in video encoders and decoders, which involve determining operational states for LIC and other prediction tools, adjusting pixel values, and enabling LIC for specific video blocks based on their positioning within coding tree units, allows for adaptive processing and improved compression.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If local illumination compensation (LIC) is applied to all video blocks, then compression ratio is improved, but encoding complexity increases

Engineering Contradiction:
Improvecompression ratioVSAvoidencoding complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent applies LIC selectively only to boundary blocks of coding tree units (CTUs) rather than uniformly to all blocks. This local quality approach enables illumination compensation where it is most needed (at boundaries where illumination changes occur) while avoiding unnecessary processing in inner blocks, thus improving compression ratio for boundary regions without proportionally increasing overall encoding complexity.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent segments the video content by distinguishing between boundary blocks and inner blocks within CTUs. This segmentation allows different LIC processing strategies to be applied to different segments, with full LIC applied to boundary blocks and inheritance or simplified processing applied to inner blocks, thereby managing encoding complexity while maintaining compression effectiveness.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If LIC parameters are derived for each video block independently, then adaptation to local conditions is improved, but processing time increases

Engineering Contradiction:
Improveadaptation to illumination conditionsVSAvoidprocessing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent performs preliminary action by deriving LIC parameters for boundary blocks first, then using these derived parameters as inheritance sources for inner blocks. This preliminary derivation at boundary locations allows inner blocks to reuse established parameters rather than performing independent derivations, reducing processing time while maintaining adaptability through the inheritance mechanism that allows parameter updates when conditions warrant.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements self-service through the parameter inheritance mechanism where inner blocks automatically acquire LIC parameters from adjacent boundary blocks without requiring independent parameter derivation. This self-service approach reduces processing time significantly, as inner blocks simply inherit and apply parameters from their neighbors rather than performing complex local derivations, while still maintaining adaptability when boundary conditions change.

Inventive Principle:
Principle #25Self-service

3Loss of information

If LIC is enabled for boundary blocks of CTU, then compression efficiency is improved, but memory access complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidmemory access complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent applies LIC and its associated memory access operations locally only to boundary blocks of CTUs rather than uniformly across all blocks. This localized approach improves compression efficiency at boundaries where illumination variations are most prominent while limiting memory access complexity to specific boundary regions, avoiding the overhead of processing inner blocks with the same memory-intensive operations.

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS11483550B2Use of virtual candidate prediction and weighted prediction in video processing
Publication Date: 2022.10.25 BYTEDANCE INC
  • US11483550B2 patent drawing
  • US11483550B2 patent drawing
  • US11483550B2 patent drawing

AI summary

A video processing method includes constructing, for a conversion between a current block of a video and a bitstream representation of the video, a motion candidate list with at least one motion candidate; and determining, based on a candidate type of a first motion candidate, whether a generalized bi-prediction (GBI) processing tool is enabled. The GBI processing tool comprises deriving a final prediction based on applying equal or unequal weights to predictions from different reference lists according to a set of weights. The method also includes performing the conversion according to the determining.